The 12 Principles of Agent Runtime Governance, Implemented
Watchlight AI Beacon
The enterprise runtime control plane for AI agents. Available now. Request a demo to see it in your environment.
A framework is easy to publish and hard to build. Anyone can write down what good governance for AI agents should look like. The test is whether there is a control plane that actually does it, on every action, in production.
We published the 12 Non-Negotiable Principles for Agent Runtime Governance as the definition of the category. This is the other half of the story: the direct mapping from each principle to the Watchlight AI Beacon capability that implements it. Eleven of the twelve are available today. One is on the roadmap, and it is marked as such.
The mapping
| Principle | Watchlight AI Beacon capability | Status |
|---|---|---|
| 1. Agent Identity | Cryptographically verifiable agent identity and registry; untrusted agents are denied | Available |
| 2. Purpose, Goals, and Intent | Intent declaration and validation, with intent expiry and justification enforced | Available |
| 3. Scoped, Time-Bound Authority | Explicit, scoped, time-bound authority grants | Available |
| 4. Deterministic Control Planes | Deterministic policy evaluation, with no model in the trust path | Available |
| 5. Plan-Act-Observe Lifecycle | Governance at the plan, act, and observe stages, before and during execution | Available |
| 6. Human-in-the-Loop | Policy-defined human checkpoints when an action crosses a risk threshold | Available |
| 7. Runtime Policy Enforcement | Per-action authorization against current, versioned policy at execution | Available |
| 8. Governed Memory and State | Governance of agent memory and persisted state | On the roadmap |
| 9. Observability and Auditability | Tamper-evident execution lineage and Agent Runtime Attestation | Available |
| 10. Safe Failure Semantics | Fail-closed enforcement with kill switches | Available |
| 11. Tool and Service Governance | Per-action governance of tools and MCP servers | Available |
| 12. Multi-Agent Coordination | Sub-agent scope attenuation and delegation-chain validation | Available |
The one on the roadmap
Principle 8, Governed Memory and State, is the governance of what an agent remembers and carries forward between tasks: the working state and long-term memory that shape its future decisions. It is on our roadmap and it is not shipping yet.
Why this matters
Two things make this mapping different from a feature list. The first is that the requirements are not ours to move. They are published, they define the category, and they hold any system that claims to provide Agent Runtime Governance to the same bar, including ours. The second is that these are not aspirations. Every capability marked available runs today against real agents in design-partner environments, deterministically, before and during each action.
That is the point of building the framework and the product together. The principles advise an enterprise on what to require. Watchlight AI Beacon is where those requirements are met.
Related reading: the 12 Non-Negotiable Principles for Agent Runtime Governance, Agent Runtime Governance Is Becoming Critical Infrastructure, and Agent Runtime Governance Is Not Agent Security.
Put runtime governance in front of every agent action
Watchlight AI Beacon is available now, fully on-premises and air-gapped. Request a demo to see it in your environment.
Agent Governance Readiness Assessment
Evaluate your governance posture against the 12 principles. Get a maturity score and roadmap.
2-3 days · Download one-pager (PDF)